Hi Frank/Malahal,

Re-opening this thread.

On 06/24/2015 04:47 AM, Frank Filz wrote:
> Note that the way Ganesha handles the epoch for clientids is structured in a 
> way that Ganesha doesn't have to be party to the details.

  There is a command line option that allows and external actor 
(presumably the startup script) to pass Ganesha a unique clientid epoch 
for each instance.

This epoch should include some kind of node identifier for a cluster, 
and some kind of time element (either a timestamp or a counter that is 
incremented each

time Ganesha starts on the node) so that each time Ganesha is started on 
a given node, a new epoch is assigned. A cluster global atomic counter 
would also work

  that would issue a new unique epoch to any Ganesha instance in the 
cluster. The epoch is 32 bits.


We have a situation where in we need to support nfs-ganesha servers in a 
clustered environment with all the nodes ntp-configured. So we cannot 
change and set different epoch values for each of them.

The only work-around we are left with is to start nfs-ganesha server 
with a delay on each node of the cluster. But that doesn't entirely 
guarantee that nfs-ganesha servers start_time will not collide 
especially during failover and restart scenarios.

So we would like to know if there is any better way to handle and 
prevent this problem.

May be ServerEpoch could contain 'boot_time + node_id/ip_addr(which can 
be provided via a config option)'.

Please share your thoughts.

Thanks,
Soumya

------------------------------------------------------------------------------
Don't Limit Your Business. Reach for the Cloud.
GigeNET's Cloud Solutions provide you with the tools and support that
you need to offload your IT needs and focus on growing your business.
Configured For All Businesses. Start Your Cloud Today.
https://www.gigenetcloud.com/
_______________________________________________
Nfs-ganesha-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/nfs-ganesha-devel

Reply via email to